This PR implements `expandSymlink` on Windows with POSIX readlink
semantics: it expands exactly one hop and returns the stored link target
without resolving the full chain.
The main design question was whether Windows symlink expansion should be
built on path-finalization APIs such as `GetFinalPathNameByHandleW`, or
on direct reparse-point inspection. Current `expandSymlink` is a
single-hop "what target is stored in this link object?" operation and
most of other ways to resolve symlinks on Windows actually try to answer
the "final true file location" question in various slightly-incompatible
ways.
The full final-path resolution on Windows is substantially more complex
than readlink and is planned as a follow-up.
## Implementation choice
Implements Windows `expandSymlink` by:
- opening the path with `FILE_FLAG_OPEN_REPARSE_POINT`
- calling `DeviceIoControl(FSCTL_GET_REPARSE_POINT)`
- parsing the reparse payload for `IO_REPARSE_TAG_SYMLINK` and
`IO_REPARSE_TAG_MOUNT_POINT`
- decoding the UTF-16 slice referenced by the payload
- returning the stored target
This is the right primitive for the API:
- does not depend on whole-path finalization
- works for both symlinks and junctions
- matches the existing Linux behaviour
`widestrs` changes allow using WideCString views without temporary
allocations.
Windows prohibits symlink creation without admin rights, so,
unfortunately, the tests are conditionally skipped by default. Manually
running `testament` in an admin console is required.
## Behaviour:
- One hop only
- Relative symlink targets are returned unchanged
- Absolute Windows targets are converted from stored NT-style prefixes
to usable Win32 forms when applicable
- Non-links, malformed payloads, and unsupported reparse tags raise
`OSError`
## Future work
Path canonicalization, i.e. "final true file location". Which is, BTW,
different from `absolutePath`, which works on paths only and doesn't hit
the underlying FS. So this needs to be an API extension.
I'd like to follow-up with this when I sort through the docs, for now
you can resolve symlinks in a loop.
---------
Co-authored-by: Andreas Rumpf <rumpf_a@web.de>
Companion to readRawData whose pointer stays valid across moves/copies
of the string. Under --strings:sso it promotes a small inline string to
its heap representation; under refc/v2 the data is already heap-resident
so it aliases readRawData.
Uniform `var string` signature on every backend so code can prepare for
--strings:sso without `when declared`.
This PR adds 3 modes to `typeof` to specify how to handle type modifiers
`var`, `sink` and `lent`.
- typeOfModCompatible
Remove or keep type modifiers in the same way as old typeof. That means
keep `sink` but remove `var` and `lent`.
- typeOfModRemoveModifier
Remove type modifiers.
- typeOfModKeepModifier
Keep type modifiers.
Related to https://github.com/nim-lang/Nim/pull/25779https://github.com/nim-lang/Nim/issues/25786
## Summary
Fixes#19782.
The `?` operator in `std/uri` was silently overwriting any query string
already present in the URI. This PR makes it append instead — which
matches the docstring ("Concatenates the query parameters") and the
natural expectation when chaining operations.
**Before:**
```nim
let u = parseUri("https://example.com/foo?existing=1") ? {"bar": "qux"}
echo $u # https://example.com/foo?bar=qux (existing=1 lost)
```
**After:**
```nim
let u = parseUri("https://example.com/foo?existing=1") ? {"bar": "qux"}
echo $u # https://example.com/foo?existing=1&bar=qux
```
## Changes
- `lib/pure/uri.nim`: fix `?` to append with `&` when a query string
already exists; add example to `runnableExamples`
- `tests/stdlib/turi.nim`: two new test cases (append to existing query,
empty params preserve existing)
- `changelog.md`: entry under Standard library changes
## Notes
I work with Claude as a co-processor. I'm 56, came to programming late,
and this is genuinely how I learn and contribute. I understand what I'm
submitting, but I didn't write it alone. If your project prefers
human-only contributions, just say so and I'll close without friction.
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: n0madgang <14005836+n0madgang@users.noreply.github.com>
## Summary
- Fixes `std/pegs` lexing for bare UTF-8 terminals such as `\i café`.
- The lexer previously stopped at the first non-ASCII byte, so
`pkTerminalIgnoreCase` never saw the full term despite its rune-aware
`fastRuneAt`/`toLower` matching.
- This now keeps non-ASCII bytes in identifier-style terminals while
ASCII non-ident characters still terminate the symbol.
## Behavior
Before: `match("CAFÉ", peg"\i café")` failed because the terminal was
lexed as `caf`.
After: `match("CAFÉ", peg"\i café")`, `match("Café", peg"\i café")`, and
`findAll` over mixed-case occurrences pass.
`std/pegs` documents `useUnicode = true` as proper UTF-8 support, and
quoted terminals already preserved the same bytes; this makes bare
terminals consistent with that path.
I did not find an existing relevant issue or PR in searches for
pegs/unicode/utf8/getSymbol/pkTerminalIgnoreCase.
fixes#25617
This pull request introduces a stricter check for parameter type
relations in the `procParamTypeRel` procedure. Specifically, it ensures
that two types are not only structurally equal but also have the same
backend type, taking type aliases into account.
Type relation checks:
*
[`compiler/sigmatch.nim`](diffhunk://#diff-251afcd01d239369019495096c187998dd6695b6457528953237a7e4a10f7138R787-R789):
In `procParamTypeRel`, added a check to ensure that if two types are
considered equal (`isEqual`), they must also have the same backend type
(using `sameBackendTypePickyAliases`). If not, the result is set to
`isNone`, preventing false positives when type aliases differ.
Adds `system.setLenUninit` for the `string` type. Allows setting length
without initializing new memory on growth.
- Required for a follow-up to #15951
- Accompanies #22767 (ref #19727) but for strings
- Expands `stdlib/tstring` with tests for `setLen` and `setLenUninit`
---------
Co-authored-by: Andreas Rumpf <araq4k@proton.me>
Adds configurable parser modes to std/parseopt module. **Take two.**
Initially solved the issue of not being able to pass arguments to short
options as you do with most everyday CLI programs, but reading the tests
made me add more features so that some of the behaviour could be changed
and here we are.
**`std/parseopt` now supports three parser modes** via an optional
`mode` parameter in `initOptParser` and `getopt`.
Three modes are provided:
- `NimMode` (default, fully backward compatible),
- `LaxMode` (POSIX-inspired with relaxed short option handling),
- `GnuMode` (stricter GNU-style conventions).
The new modes are marked as experimental in the documentation.
The parser behaviour is controlled by a new `ParserRules` enum, which
provides granular feature flags that modes are built from. This makes it
possible for users with specific requirements to define custom rule sets
by importing private symbols, this is mentioned but clearly marked as
unsupported.
**Backward compatibility:**
The default mode preserves existing behaviour completely, with a single
exception: `allowWhitespaceAfterColon` is deprecated.
Now, `allowWhitespaceAfterColon` doesn't make much sense as a single
tuning knob. The `ParserRule.prSepAllowDelimAfter` controls this now.
As `allowWhitespaceAfterColon` had a default, most calls never mention
it so they will silently migrate to the new `initOptParser` overload. To
cover cases when the proc param was used at call-site, I added an
overload, which modifies the default parser mode to reflect the required
`allowWhitespaceAfterColon` value. Should be all smooth for most users,
except the deprecation warning.
The only thing I think can be classified as the breaking change is a
surprising **bug** of the old parser:
```nim
let p = initOptParser("-n 10 -m20 -k= 30 -40", shortNoVal = {'v'})
# ^-disappears
```
This is with the aforementioned `allowWhitespaceAfterColon` being true
by default, of course. In this case the `30` token is skipped
completely. I don't think that's right, so it's fixed.
Things I still don't like about how the old parser and the new default
mode behave:
1. **Parser behaviour is controlled by an emptiness of two containers**.
This is an interesting approach. It's also made more interesting because
the `shortNoVal`/`longNoVal` control both the namesakes, but *and also
how their opposites (value-taking opts) work*.
---
**Edit:**
2. `shortNoVal` is not mandatory:
```nim
let p = initOptParser(@["-a=foo"], shortNoVal = {'a'})
# Nim, Lax parses as: (cmdShortOption, "a", "foo")
# GnuMode parses as: (cmdShortOption, "a", "=foo")
```
In this case, even though the user specified `a` as no no-val, parser
ignores it, relying only on the syntax to decide the kind of the
argument. This is especially problematic with the modes that don't use
the rule `prShortAllowSep` (GnuMode), in this case the provided input is
twice invalid, regardless of the `shortNoVal`.
With the current parser architecture, parsing it this way **is
inevitable**, though. We don't have any way to signal the error state
detected with the input, so the user is expected to validate the input
for mistakes.
Bundling positional arguments is nonsensical and short option can't use
the separator character, so `[cmd "a", arg "=foo"]` and `[cmd "a", cmd
"=", cmd "f"...]` are both out of the question **and** would complicate
validating, requiring keeping track of a previous argument. Hope I'm
clear enough on the issue.
**Future work:**
1. Looks like the new modes are already usable, but from the discussions
elsewhere it looks like we might want to support special-casing
multi-digit short options (`-XX..`) to allow numerical options greater
than 9. This complicates bundling, though, so requires a bit of thinking
through.
2. Signaling error state?
---------
Co-authored-by: Andreas Rumpf <araq4k@proton.me>
Implements #21928
Adds a `--raw` (since thats what the original issue used, suggestions
welcome) switch which stops the jsondoc gen from rendering rst/markdown.
Implemented by making `genComment` check if it needs to return the raw
string or not. This required switching the related procs to using
`Option` to handle how `nil` values were returned before. The `nil`
returns were eventually ignored so just ignoring `none(T)` has the same
effect.
Doesn't support `runnableExamples` since jsondocs doesn't support them
either
The `hasValuelessStatics` function in `semtypinst.nim` only checked for
`tyStatic`, missing `tyTypeDesc(tyGenericParam)`. This caused
`sizeof(T)` inside a typedesc template called from a generic type's
`when` clause to error with "'sizeof' requires '.importc' types to be
'.completeStruct'".
The fix adds a check for `tyTypeDesc` wrapping `tyGenericParam`,
recognizing it as an unresolved generic parameter that needs resolution
before evaluation.
Also documents the `completeStruct` pragma in the manual.
fixes#24998
Basically it retraces back to the situation before
https://github.com/nim-lang/Nim/pull/18366 and
https://github.com/nim-lang/Nim/pull/18362, i.e.
```nim
import fuzz/a
import fuzz/a
```
```nim
import fuzz/a
from buzz/a
```
```nim
import fuzz/a except nil
from fuzz/a import addInt
```
All of these cases are now flagged as invalid and triggers a
redefinition error, i.e., each module name importing is treated as
consistent as the symbol definition
kinda annoying for importing/exporting with `when conditions` though
ref https://github.com/nim-lang/Nim/issues/18762https://github.com/nim-lang/Nim/issues/20907
```nim
from std/strutils import toLower
when not defined(js):
from std/strutils import toUpper
```
fixes#25007
```nim
proc setLengthSeqUninit(s: PGenericSeq, typ: PNimType, newLen: int, isTrivial: bool): PGenericSeq {.
compilerRtl.} =
```
In this added function, only the line `zeroMem(dataPointer(result,
elemAlign, elemSize, newLen), (result.len-%newLen) *% elemSize)` is
removed from `proc setLengthSeqV2` when enlarging a sequence.
JS and VM versions simply use `setLen`.
todo: We can also give a deprecation message for `ltPtr`/`lePtr`
matching for cstring in `magicsAfterOverloadResolution`
follow up https://github.com/nim-lang/Nim/pull/24942
Implements #24569
Adds `--stdinfile` flag for specifying the file to use in place of
`stdinfile.nim` in error messages. Will enable easier integration of
tooling with nim check
implements https://github.com/nim-lang/RFCs/issues/557
It inserts defect handing into a bare except branch
```nim
try:
raiseAssert "test"
except:
echo "nope"
```
=>
```nim
try:
raiseAssert "test"
except:
# New behaviov, now well-defined: **never** catches the assert, regardless of panic mode
raiseDefect()
echo "nope"
```
In this way, `except` still catches foreign exceptions, but panics on
`Defect`. Probably when Nim has `except {.foreign.}`, we can extend
`raiseDefect` to foreign exceptions as well. That's supposed to be a
small use case anyway.
`--legacy:noPanicOnExcept` is provided for a transition period.
- Allows using with `--experimental:strictFuncs`
- `{.cast(noSideEffect).}:` inside the proc was required to mutate
`s.len`, same as used in `newSeqImpl`.
- Removed now unnecessary `noSideEffect` casts in `system.nim`
-
Closes#24811
Co-authored-by: ringabout <43030857+ringabout@users.noreply.github.com>
- `system.substr` now uses `copymem` when available, introducing a small
template for nimvm detection (#12517#12518)
- Docs are updated to clarify behaviour on out-of-bounds input
- Runnable examples cover more edge cases and do not repeat between
overloads
- Docs now explain the difference between overloads
What bothers me is that the `substr*(a: openArray[char]): string =`
which was added by @beef331 is practically an implementation of #14810,
which is just a conversion from `openArray` to `string` but somehow it
ended up being a `substr` overload, even though its behaviour is totally
different, _the "substringing" is performed by a previous step_
(conversion to openArray) and the bounds are not checked. I'm not sure
it's that great for overloads to differ in subtle ways so much.
What are the cases that `substr` covers now, that prohibit renaming it
to `toString` (or something like that)?
Multiple replacements based on character sets in a single pass. Useful
for string sanitation. Follows existing `multiReplace` semantics.
Note: initially copied the substring version logic with a `while` and a
named block break, but Godbolt showed it had produced slightly larger
assembly using higher registers than the final version.
- [x] Tests
- [x] changelog.md
fixes#23114
As in https://github.com/nim-lang/Nim/pull/22074, expressions in
bracketed emit are strictly typechecked, this PR applies the same check
for symbols in asm statements in order to keep them consistent.
closes https://github.com/nim-lang/RFCs/issues/380, fixes#4773, fixes
#14729, fixes#16755, fixes#18150, fixes#22984, refs #11167 (only some
comments fixed), refs #12620 (needs tiny workaround)
The compiler gains a concept of root "nominal" types (i.e. objects,
enums, distincts, direct `Foo = ref object`s, generic versions of all of
these). Exported top-level routines in the same module as the nominal
types that their parameter types derive from (i.e. with
`var`/`sink`/`typedesc`/generic constraints) are considered attached to
the respective type, as the RFC states. This happens for every argument
regardless of placement.
When a call is overloaded and overload matching starts, for all
arguments in the call that already have a type, we add any operation
with the same name in the scope of the root nominal type of each
argument (if it exists) to the overload match. This also happens as
arguments gradually get typed after every overload match. This restricts
the considered overloads to ones attached to the given arguments, as
well as preventing `untyped` arguments from being forcefully typed due
to unrelated overloads. There are some caveats:
* If no overloads with a name are in scope, type bound ops are not
triggered, i.e. if `foo` is not declared, `foo(x)` will not consider a
type bound op for `x`.
* If overloads in scope do not have enough parameters up to the argument
which needs its type bound op considered, then type bound ops are also
not added. For example, if only `foo()` is in scope, `foo(x)` will not
consider a type bound op for `x`.
In the cases of "generic interfaces" like `hash`, `$`, `items` etc. this
is not really a problem since any code using it will have at least one
typed overload imported. For arbitrary versions of these though, as in
the test case for #12620, a workaround is to declare a temporary
"template" overload that never matches:
```nim
# neither have to be exported, just needed for any use of `foo`:
type Placeholder = object
proc foo(_: Placeholder) = discard
```
I don't know what a "proper" version of this could be, maybe something
to do with the new concepts.
Possible directions:
A limitation with the proposal is that parameters like `a: ref Foo` are
not attached to any type, even if `Foo` is nominal. Fixing this for just
`ptr`/`ref` would be a special case, parameters like `seq[Foo]` would
still not be attached to `Foo`. We could also skip any *structural* type
but this could produce more than one nominal type, i.e. `(Foo, Bar)`
(not that this is hard to implement, it just might be unexpected).
Converters do not use type bound ops, they still need to be in scope to
implicitly convert. But maybe they could also participate in the nominal
type consideration: if `Generic[T] = distinct T` has a converter to `T`,
both `Generic` and `T` can be considered as nominal roots.
The other restriction in the proposal, being in the same scope as the
nominal type, could maybe be worked around by explicitly attaching to
the type, i.e.: `proc foo(x: T) {.attach: T.}`, similar to class
extensions in newer OOP languages. The given type `T` needs to be
obtainable from the type of the given argument `x` however, i.e.
something like `proc foo(x: ref T) {.attach: T.}` doesn't work to fix
the `ref` issue since the compiler never obtains `T` from a given `ref
T` argument. Edit: Since the module is queried now, this is likely not
possible.
---------
Co-authored-by: Andreas Rumpf <rumpf_a@web.de>
closes https://github.com/nim-lang/RFCs/issues/554
Adds a symmetric difference operation to the language bitset type. This
maps to a simple `xor` operation on the backend and thus is likely
faster than the current alternatives, namely `(a - b) + (b - a)` or `a +
b - a * b`. The compiler VM implementation of bitsets already
implemented this via `symdiffSets` but it was never used.
The standalone binary operation is added to `setutils`, named
`symmetricDifference` in line with [hash
sets](https://nim-lang.org/docs/sets.html#symmetricDifference%2CHashSet%5BA%5D%2CHashSet%5BA%5D).
An operator version `-+-` and an in-place version like `toggle` as
described in the RFC are also added, implemented as trivial sugar.
I have added a new overload of `^` for float exponents.
Is two overloads for `float32` and `float64` better than just one
overload with `SomeFloat` type ?
I guess this would not work with `SomeFloat`, as `pow` is not defined
for `float`.
Another remark. Maybe we should catch exponents with 0.5 and call `sqrt`
instead ?
---------
Co-authored-by: Clay Sweetser <Varriount@users.noreply.github.com>
Co-authored-by: metagn <metagngn@gmail.com>
refs #24207
The `-d:nimUseCAtomics` flag added in #24207 is now inverted and made
into `-d:nimUseCppAtomics`, which means C++ atomics are only enabled
with the define. This flag is now also documented and tested.
alternative to #24101#23892 changed the opensym experimental switch so that it has to be
enabled in the context of the generic/template declarations capturing
the symbols, not the context of the instantiation of the
generics/templates. This was to be in line with where the compiler gives
the warnings and changes behavior in a potentially breaking way.
However `results` [depends on the old
behavior](71d404b314/results.nim (L1428)),
so that the callers of the macros provided by results always take
advantage of the opensym behavior. To accomodate this, we change the
behavior of the old experimental option that `results` uses,
`genericsOpenSym`, so that ignores the information of whether or not
symbols are intentionally opened and always gives the opensym behavior
as long as it's enabled at instantiation time. This should keep
`results` working as is. However this differs from the normal opensym
switch in that it doesn't generate `nnkOpenSym`.
Before it was just a generics-only version of `openSym` along with
`templateOpenSym` which was only for templates. So `templateOpenSym` is
removed along with this change, but no one appears to have used it.
fixes#5395
Previously values of `const` statements used the same scope as the
`const` statement itself, meaning variables could be declared inside
them and referred to in other statements in the same block. Now each
`const` value opens its own scope, so any variable declared in the value
of a constant can only be accessed for that constant.
We could change this to open a new scope for the `const` *section*
rather than each constant, so the variables can be used in other
constants, but I'm not sure if this is sound.
refs #21682, refs #24038
The `rangeBase` typetrait added in #21682 which gives the base type of a
range type is now added publicly to `typetraits`. Previously it was only
privately used in `repr_v2` and in `enumutils` since #24052
(coincidentally I didn't see this until now). This is part of an effort
to make range types easier to work with in generics, as mentioned in
#24038. Its use combined with #24037 is also tested.
The condition for the "enum to enum conversion" warning is now also
restricted to conversions between different enum base types, i.e.
conversion between an enum type and a range type of itself doesn't give
a warning. I put this in this PR since the test gave the warning and so
works as a regression test.
fixes#15314, fixes#24002
The OpenSym behavior first added to generics in #23091 now also applies
to templates, since templates can also capture symbols that are meant to
be replaced by local symbols if the context imports symbols with the
same name, as in the issue #24002. The experimental switch
`templateOpenSym` is added to enable this behavior for templates only,
and the experimental switch `openSym` is added to enable it for both
templates and generics, and the documentation now mainly mentions this
switch.
Additionally the logic for `nkOpenSymChoice` nodes that were previously
wrapped in `nkOpenSym` now apply to all `nkOpenSymChoice` nodes, and so
these nodes aren't wrapped in `nkOpenSym` anymore. This means
`nkOpenSym` can only have children of kind `nkSym` again, so it is more
in line with the structure of symchoice nodes. As for why they aren't
merged with `nkOpenSymChoice` nodes yet, we need some way to signal that
the node shouldn't become ambiguous if other options exist at
instantiation time, we already captured a symbol at the beginning and
another symbol can only replace it if it's closer in scope and
unambiguous.
fixes#23898, supersedes #23966 and #23990
Since #20631 ambiguous type symbols in templates are rejected outright,
now we generate a symchoice for type nodes if they're ambiguous, a
generalization of what was done in #22375. This is done for generics as
well. Symchoices also handle type symbols better now, ensuring their
type is a `typedesc` type; this probably isn't necessary for everything
to work but it makes the logic more robust.
Similar to #23989, we have to prepare for the fact that ambiguous type
symbols behave differently than normal type symbols and either error
normally or relegate to other routine symbols if the symbol is being
called. Generating a symchoice emulates this behavior, `semExpr` will
find the type symbol first, but since the symchoice has other symbols,
it will count as an ambiguous type symbol.
I know it seems spammy to carry around an ambiguity flag everywhere, but
in the future when we have something like #23104 we could just always
generate a symchoice, and the symchoice itself would carry the info of
whether the first symbol was ambiguous. But this could harm compiler
performance/memory use, it might be better to generate it only when we
have to, which in the case of type symbols is only when they're
ambiguous.
